]> git.ipfire.org Git - thirdparty/mdadm.git/blob - mdadm.h
Config: use better device names for "DEVICES container"
[thirdparty/mdadm.git] / mdadm.h
1 /*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
22 * Email: <neilb@suse.de>
23 */
24
25 #define _GNU_SOURCE
26 #define _FILE_OFFSET_BITS 64
27 #include <unistd.h>
28 #ifdef __GLIBC__
29 extern __off64_t lseek64 __P ((int __fd, __off64_t __offset, int __whence));
30 #elif !defined(lseek64)
31 # if defined(__NO_STAT64) || __WORDSIZE != 32
32 # define lseek64 lseek
33 # endif
34 #endif
35
36 #include <sys/types.h>
37 #include <sys/stat.h>
38 #include <stdlib.h>
39 #include <time.h>
40 #include <sys/time.h>
41 #include <getopt.h>
42 #include <fcntl.h>
43 #include <stdio.h>
44 #include <errno.h>
45 #include <string.h>
46 #include <syslog.h>
47 #ifdef __dietlibc__
48 #include <strings.h>
49 /* dietlibc has deprecated random and srandom!! */
50 #define random rand
51 #define srandom srand
52 #endif
53
54 #include <linux/kdev_t.h>
55 /*#include <linux/fs.h> */
56 #include <sys/mount.h>
57 #include <asm/types.h>
58 #include <sys/ioctl.h>
59 #define MD_MAJOR 9
60 #define MdpMinorShift 6
61
62 #ifndef BLKGETSIZE64
63 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
64 #endif
65
66 #define DEFAULT_CHUNK 512
67 #define DEFAULT_BITMAP_CHUNK 4096
68 #define DEFAULT_BITMAP_DELAY 5
69 #define DEFAULT_MAX_WRITE_BEHIND 256
70
71 /* MAP_DIR should be somewhere that persists across the pivotroot
72 * from early boot to late boot.
73 * /run seems to have emerged as the best standard.
74 */
75 #ifndef MAP_DIR
76 #define MAP_DIR "/run/mdadm"
77 #endif /* MAP_DIR */
78 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
79 * want something other than the default of "map"
80 */
81 #ifndef MAP_FILE
82 #define MAP_FILE "map"
83 #endif /* MAP_FILE */
84 /* MDMON_DIR is where pid and socket files used for communicating
85 * with mdmon normally live. Best is /var/run/mdadm as
86 * mdmon is needed at early boot then it needs to write there prior
87 * to /var/run being mounted read/write, and it also then needs to
88 * persist beyond when /var/run is mounter read-only. So, to be
89 * safe, the default is somewhere that is read/write early in the
90 * boot process and stays up as long as possible during shutdown.
91 */
92 #ifndef MDMON_DIR
93 #define MDMON_DIR "/run/mdadm"
94 #endif /* MDMON_DIR */
95
96 /* FAILED_SLOTS is where to save files storing recent removal of array
97 * member in order to allow future reuse of disk inserted in the same
98 * slot for array recovery
99 */
100 #ifndef FAILED_SLOTS_DIR
101 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
102 #endif /* FAILED_SLOTS */
103
104 #include "md_u.h"
105 #include "md_p.h"
106 #include "bitmap.h"
107 #include "msg.h"
108
109 #include <endian.h>
110 /* Redhat don't like to #include <asm/byteorder.h>, and
111 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
112 * and there is no standard conversion function so... */
113 /* And dietlibc doesn't think byteswap is ok, so.. */
114 /* #include <byteswap.h> */
115 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
116 ((x) & 0xff00U) >> 8)
117 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
118 ((x) & 0xff000000U) >> 24 | \
119 ((x) & 0x0000ff00U) << 8 | \
120 ((x) & 0x00ff0000U) >> 8)
121 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
122 ((x) & 0xff00000000000000ULL) >> 56 | \
123 ((x) & 0x000000000000ff00ULL) << 40 | \
124 ((x) & 0x00ff000000000000ULL) >> 40 | \
125 ((x) & 0x0000000000ff0000ULL) << 24 | \
126 ((x) & 0x0000ff0000000000ULL) >> 24 | \
127 ((x) & 0x00000000ff000000ULL) << 8 | \
128 ((x) & 0x000000ff00000000ULL) >> 8)
129
130 #if !defined(__KLIBC__)
131 #if BYTE_ORDER == LITTLE_ENDIAN
132 #define __cpu_to_le16(_x) (_x)
133 #define __cpu_to_le32(_x) (_x)
134 #define __cpu_to_le64(_x) (_x)
135 #define __le16_to_cpu(_x) (_x)
136 #define __le32_to_cpu(_x) (_x)
137 #define __le64_to_cpu(_x) (_x)
138
139 #define __cpu_to_be16(_x) bswap_16(_x)
140 #define __cpu_to_be32(_x) bswap_32(_x)
141 #define __cpu_to_be64(_x) bswap_64(_x)
142 #define __be16_to_cpu(_x) bswap_16(_x)
143 #define __be32_to_cpu(_x) bswap_32(_x)
144 #define __be64_to_cpu(_x) bswap_64(_x)
145 #elif BYTE_ORDER == BIG_ENDIAN
146 #define __cpu_to_le16(_x) bswap_16(_x)
147 #define __cpu_to_le32(_x) bswap_32(_x)
148 #define __cpu_to_le64(_x) bswap_64(_x)
149 #define __le16_to_cpu(_x) bswap_16(_x)
150 #define __le32_to_cpu(_x) bswap_32(_x)
151 #define __le64_to_cpu(_x) bswap_64(_x)
152
153 #define __cpu_to_be16(_x) (_x)
154 #define __cpu_to_be32(_x) (_x)
155 #define __cpu_to_be64(_x) (_x)
156 #define __be16_to_cpu(_x) (_x)
157 #define __be32_to_cpu(_x) (_x)
158 #define __be64_to_cpu(_x) (_x)
159 #else
160 # error "unknown endianness."
161 #endif
162 #endif /* __KLIBC__ */
163
164 /*
165 * min()/max()/clamp() macros that also do
166 * strict type-checking.. See the
167 * "unnecessary" pointer comparison.
168 */
169 #define min(x, y) ({ \
170 typeof(x) _min1 = (x); \
171 typeof(y) _min2 = (y); \
172 (void) (&_min1 == &_min2); \
173 _min1 < _min2 ? _min1 : _min2; })
174
175 #define max(x, y) ({ \
176 typeof(x) _max1 = (x); \
177 typeof(y) _max2 = (y); \
178 (void) (&_max1 == &_max2); \
179 _max1 > _max2 ? _max1 : _max2; })
180
181 /* general information that might be extracted from a superblock */
182 struct mdinfo {
183 mdu_array_info_t array;
184 mdu_disk_info_t disk;
185 __u64 events;
186 int uuid[4];
187 char name[33];
188 unsigned long long data_offset;
189 unsigned long long new_data_offset;
190 unsigned long long component_size; /* same as array.size, except in
191 * sectors and up to 64bits.
192 */
193 unsigned long long custom_array_size; /* size for non-default sized
194 * arrays (in sectors)
195 */
196 #define NO_RESHAPE 0
197 #define VOLUME_RESHAPE 1
198 #define CONTAINER_RESHAPE 2
199 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
200 int reshape_active;
201 unsigned long long reshape_progress;
202 int recovery_blocked; /* for external metadata it
203 * indicates that there is
204 * reshape in progress in
205 * container,
206 * for native metadata it is
207 * reshape_active field mirror
208 */
209 /* During reshape we can sometimes change the data_offset to avoid
210 * over-writing still-valid data. We need to know if there is space.
211 * So getinfo_super will fill in space_before and space_after in sectors.
212 * data_offset can be increased or decreased by this amount.
213 */
214 unsigned long long space_before, space_after;
215 union {
216 unsigned long long resync_start; /* per-array resync position */
217 unsigned long long recovery_start; /* per-device rebuild position */
218 #define MaxSector (~0ULL) /* resync/recovery complete position */
219 };
220 long bitmap_offset; /* 0 == none, 1 == a file */
221 unsigned long safe_mode_delay; /* ms delay to mark clean */
222 int new_level, delta_disks, new_layout, new_chunk;
223 int errors;
224 unsigned long cache_size; /* size of raid456 stripe cache*/
225 int mismatch_cnt;
226 char text_version[50];
227
228 int container_member; /* for assembling external-metatdata arrays
229 * This is to be used internally by metadata
230 * handler only */
231 int container_enough; /* flag external handlers can set to
232 * indicate that subarrays have not enough (-1),
233 * enough to start (0), or all expected disks (1) */
234 char sys_name[20];
235 struct mdinfo *devs;
236 struct mdinfo *next;
237
238 /* Device info for mdmon: */
239 int recovery_fd;
240 int state_fd;
241 #define DS_FAULTY 1
242 #define DS_INSYNC 2
243 #define DS_WRITE_MOSTLY 4
244 #define DS_SPARE 8
245 #define DS_BLOCKED 16
246 #define DS_REMOVE 1024
247 #define DS_UNBLOCK 2048
248 int prev_state, curr_state, next_state;
249
250 };
251
252 struct createinfo {
253 int uid;
254 int gid;
255 int autof;
256 int mode;
257 int symlinks;
258 int names;
259 struct supertype *supertype;
260 };
261
262 #define Name "mdadm"
263
264 enum mode {
265 ASSEMBLE=1,
266 BUILD,
267 CREATE,
268 MANAGE,
269 MISC,
270 MONITOR,
271 GROW,
272 INCREMENTAL,
273 AUTODETECT,
274 mode_count
275 };
276
277 extern char short_options[];
278 extern char short_bitmap_options[];
279 extern char short_bitmap_auto_options[];
280 extern struct option long_options[];
281 extern char Version[], Usage[], Help[], OptionHelp[],
282 *mode_help[],
283 Help_create[], Help_build[], Help_assemble[], Help_grow[],
284 Help_incr[],
285 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
286
287 /* for option that don't have short equivilents, we assign arbitrary
288 * numbers later than any 'short' character option.
289 */
290 enum special_options {
291 AssumeClean = 300,
292 BitmapChunk,
293 WriteBehind,
294 ReAdd,
295 NoDegraded,
296 Sparc22,
297 BackupFile,
298 HomeHost,
299 AutoHomeHost,
300 Symlinks,
301 AutoDetect,
302 Waitclean,
303 DetailPlatform,
304 KillSubarray,
305 UpdateSubarray,
306 IncrementalPath,
307 NoSharing,
308 HelpOptions,
309 Brief,
310 ManageOpt,
311 Add,
312 Remove,
313 Fail,
314 Replace,
315 With,
316 MiscOpt,
317 WaitOpt,
318 ConfigFile,
319 ChunkSize,
320 WriteMostly,
321 Layout,
322 Auto,
323 Force,
324 SuperMinor,
325 EMail,
326 ProgramOpt,
327 Increment,
328 Fork,
329 Bitmap,
330 RebuildMapOpt,
331 InvalidBackup,
332 UdevRules,
333 FreezeReshape,
334 Continue,
335 OffRootOpt,
336 Prefer,
337 KillOpt,
338 DataOffset,
339 ExamineBB,
340 Dump,
341 Restore,
342 };
343
344 enum prefix_standard {
345 JEDEC,
346 IEC
347 };
348
349 /* structures read from config file */
350 /* List of mddevice names and identifiers
351 * Identifiers can be:
352 * uuid=128-hex-uuid
353 * super-minor=decimal-minor-number-from-superblock
354 * devices=comma,separated,list,of,device,names,with,wildcards
355 *
356 * If multiple fields are present, the intersection of all matching
357 * devices is considered
358 */
359 #define UnSet (0xfffe)
360 struct mddev_ident {
361 char *devname;
362
363 int uuid_set;
364 int uuid[4];
365 char name[33];
366
367 int super_minor;
368
369 char *devices; /* comma separated list of device
370 * names with wild cards
371 */
372 int level;
373 int raid_disks;
374 int spare_disks;
375 struct supertype *st;
376 int autof; /* 1 for normal, 2 for partitioned */
377 char *spare_group;
378 char *bitmap_file;
379 int bitmap_fd;
380
381 char *container; /* /dev/whatever name of container, or
382 * uuid of container. You would expect
383 * this to be the 'devname' or UUID
384 * of some other entry.
385 */
386 char *member; /* subarray within a container */
387
388 struct mddev_ident *next;
389 union {
390 /* fields needed by different users of this structure */
391 int assembled; /* set when assembly succeeds */
392 };
393 };
394
395 struct context {
396 int readonly;
397 int runstop;
398 int verbose;
399 int brief;
400 int force;
401 char *homehost;
402 int require_homehost;
403 char *prefer;
404 int export;
405 int test;
406 char *subarray;
407 char *update;
408 int scan;
409 int SparcAdjust;
410 int autof;
411 int delay;
412 int freeze_reshape;
413 char *backup_file;
414 int invalid_backup;
415 };
416
417 struct shape {
418 int raiddisks;
419 int sparedisks;
420 int level;
421 int layout;
422 char *layout_str;
423 int chunk;
424 int bitmap_chunk;
425 char *bitmap_file;
426 int assume_clean;
427 int write_behind;
428 unsigned long long size;
429 };
430
431 /* List of device names - wildcards expanded */
432 struct mddev_dev {
433 char *devname;
434 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
435 * 'A' for re_add.
436 * Not set for names read from .config
437 */
438 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
439 int used; /* set when used */
440 long long data_offset;
441 struct mddev_dev *next;
442 };
443
444 typedef struct mapping {
445 char *name;
446 int num;
447 } mapping_t;
448
449 struct mdstat_ent {
450 char *dev;
451 char devnm[32];
452 int active;
453 char *level;
454 char *pattern; /* U or up, _ for down */
455 int percent; /* -1 if no resync */
456 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
457 int devcnt;
458 int raid_disks;
459 char * metadata_version;
460 struct dev_member {
461 char *name;
462 struct dev_member *next;
463 } *members;
464 struct mdstat_ent *next;
465 };
466
467 extern struct mdstat_ent *mdstat_read(int hold, int start);
468 extern void free_mdstat(struct mdstat_ent *ms);
469 extern void mdstat_wait(int seconds);
470 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
471 extern int mddev_busy(char *devnm);
472 extern struct mdstat_ent *mdstat_by_component(char *name);
473 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, char *container);
474
475 struct map_ent {
476 struct map_ent *next;
477 char devnm[32];
478 char metadata[20];
479 int uuid[4];
480 int bad;
481 char *path;
482 };
483 extern int map_update(struct map_ent **mpp, char *devnm, char *metadata,
484 int uuid[4], char *path);
485 extern void map_remove(struct map_ent **map, char *devnm);
486 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
487 #ifdef MDASSEMBLE
488 static inline struct map_ent *map_by_devnm(struct map_ent **map, char *name)
489 {
490 return NULL;
491 }
492 static inline void map_free(struct map_ent *map)
493 {
494 }
495 #else
496 extern struct map_ent *map_by_devnm(struct map_ent **map, char *devnm);
497 extern void map_free(struct map_ent *map);
498 #endif
499 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
500 extern void map_read(struct map_ent **melp);
501 extern int map_write(struct map_ent *mel);
502 extern void map_delete(struct map_ent **mapp, char *devnm);
503 extern void map_add(struct map_ent **melp,
504 char *devnm, char *metadata, int uuid[4], char *path);
505 extern int map_lock(struct map_ent **melp);
506 extern void map_unlock(struct map_ent **melp);
507 extern void map_fork(void);
508
509 /* various details can be requested */
510 enum sysfs_read_flags {
511 GET_LEVEL = (1 << 0),
512 GET_LAYOUT = (1 << 1),
513 GET_COMPONENT = (1 << 2),
514 GET_CHUNK = (1 << 3),
515 GET_CACHE = (1 << 4),
516 GET_MISMATCH = (1 << 5),
517 GET_VERSION = (1 << 6),
518 GET_DISKS = (1 << 7),
519 GET_DEGRADED = (1 << 8),
520 GET_SAFEMODE = (1 << 9),
521 GET_BITMAP_LOCATION = (1 << 10),
522
523 GET_DEVS = (1 << 20), /* gets role, major, minor */
524 GET_OFFSET = (1 << 21),
525 GET_SIZE = (1 << 22),
526 GET_STATE = (1 << 23),
527 GET_ERROR = (1 << 24),
528 };
529
530 /* If fd >= 0, get the array it is open on,
531 * else use devnm.
532 */
533 extern int sysfs_open(char *devnm, char *devname, char *attr);
534 extern void sysfs_init(struct mdinfo *mdi, int fd, char *devnm);
535 extern void sysfs_free(struct mdinfo *sra);
536 extern struct mdinfo *sysfs_read(int fd, char *devnm, unsigned long options);
537 extern int sysfs_attr_match(const char *attr, const char *str);
538 extern int sysfs_match_word(const char *word, char **list);
539 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
540 char *name, char *val);
541 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
542 char *name, unsigned long long val);
543 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
544 char *name, long long val);
545 extern int sysfs_uevent(struct mdinfo *sra, char *event);
546 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
547 char *name);
548 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
549 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
550 char *name, unsigned long long *val);
551 extern int sysfs_fd_get_two(int fd, unsigned long long *v1, unsigned long long *v2);
552 extern int sysfs_get_two(struct mdinfo *sra, struct mdinfo *dev,
553 char *name, unsigned long long *v1, unsigned long long *v2);
554 extern int sysfs_fd_get_str(int fd, char *val, int size);
555 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
556 char *name);
557 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
558 char *name, char *val, int size);
559 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
560 extern int sysfs_set_array(struct mdinfo *info, int vers);
561 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
562 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
563 extern int sysfs_unique_holder(char *devnm, long rdev);
564 extern int sysfs_freeze_array(struct mdinfo *sra);
565 extern int sysfs_wait(int fd, int *msec);
566 extern int load_sys(char *path, char *buf);
567 extern int reshape_prepare_fdlist(char *devname,
568 struct mdinfo *sra,
569 int raid_disks,
570 int nrdisks,
571 unsigned long blocks,
572 char *backup_file,
573 int *fdlist,
574 unsigned long long *offsets);
575 extern void reshape_free_fdlist(int *fdlist,
576 unsigned long long *offsets,
577 int size);
578 extern int reshape_open_backup_file(char *backup,
579 int fd,
580 char *devname,
581 long blocks,
582 int *fdlist,
583 unsigned long long *offsets,
584 int restart);
585 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
586 unsigned int ndata, unsigned int odata);
587
588 extern int save_stripes(int *source, unsigned long long *offsets,
589 int raid_disks, int chunk_size, int level, int layout,
590 int nwrites, int *dest,
591 unsigned long long start, unsigned long long length,
592 char *buf);
593 extern int restore_stripes(int *dest, unsigned long long *offsets,
594 int raid_disks, int chunk_size, int level, int layout,
595 int source, unsigned long long read_offset,
596 unsigned long long start, unsigned long long length,
597 char *src_buf);
598
599 #ifndef Sendmail
600 #define Sendmail "/usr/lib/sendmail -t"
601 #endif
602
603 #define SYSLOG_FACILITY LOG_DAEMON
604
605 extern char *map_num(mapping_t *map, int num);
606 extern int map_name(mapping_t *map, char *name);
607 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
608
609 extern char *map_dev_preferred(int major, int minor, int create,
610 char *prefer);
611 static inline char *map_dev(int major, int minor, int create)
612 {
613 return map_dev_preferred(major, minor, create, NULL);
614 }
615
616 struct active_array;
617 struct metadata_update;
618
619 /* 'struct reshape' records the intermediate states of
620 * a general reshape.
621 * The starting geometry is converted to the 'before' geometry
622 * by at most an atomic level change. They could be the same.
623 * Similarly the 'after' geometry is converted to the final
624 * geometry by at most a level change.
625 * Note that 'before' and 'after' must have the same level.
626 * 'blocks' is the minimum number of sectors for a reshape unit.
627 * This will be a multiple of the stripe size in each of the
628 * 'before' and 'after' geometries.
629 * If 'blocks' is 0, no restriping is necessary.
630 * 'min_offset_change' is the minimum change to data_offset to
631 * allow the reshape to happen. It is at least the larger of
632 * the old and new chunk sizes, and typically the same as 'blocks'
633 * divided by number of data disks.
634 */
635 struct reshape {
636 int level;
637 int parity; /* number of parity blocks/devices */
638 struct {
639 int layout;
640 int data_disks;
641 } before, after;
642 unsigned long long backup_blocks;
643 unsigned long long min_offset_change;
644 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
645 unsigned long long new_size; /* New size of array in sectors */
646 };
647
648 /* A superswitch provides entry point the a metadata handler.
649 *
650 * The superswitch primarily operates on some "metadata" that
651 * is accessed via the 'supertype'.
652 * This metadata has one of three possible sources.
653 * 1/ It is read from a single device. In this case it may not completely
654 * describe the array or arrays as some information might be on other
655 * devices.
656 * 2/ It is read from all devices in a container. In this case all
657 * information is present.
658 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
659 * be complete once enough ->add_to_super calls have completed.
660 *
661 * When creating an array inside a container, the metadata will be
662 * formed by a combination of 2 and 3. The metadata or the array is read,
663 * then new information is added.
664 *
665 * The metadata must sometimes have a concept of a 'current' array
666 * and a 'current' device.
667 * The 'current' array is set by init_super to be the newly created array,
668 * or is set by super_by_fd when it finds it is looking at an array inside
669 * a container.
670 *
671 * The 'current' device is either the device that the metadata was read from
672 * in case 1, or the last device added by add_to_super in case 3.
673 * Case 2 does not identify a 'current' device.
674 */
675 extern struct superswitch {
676
677 /* Used to report details of metadata read from a component
678 * device. ->load_super has been called.
679 */
680 void (*examine_super)(struct supertype *st, char *homehost);
681 void (*brief_examine_super)(struct supertype *st, int verbose);
682 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
683 void (*export_examine_super)(struct supertype *st);
684 int (*examine_badblocks)(struct supertype *st, int fd, char *devname);
685 int (*copy_metadata)(struct supertype *st, int from, int to);
686
687 /* Used to report details of an active array.
688 * ->load_super was possibly given a 'component' string.
689 */
690 void (*detail_super)(struct supertype *st, char *homehost);
691 void (*brief_detail_super)(struct supertype *st);
692 void (*export_detail_super)(struct supertype *st);
693
694 /* Optional: platform hardware / firmware details */
695 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
696 int (*export_detail_platform)(int verbose, char *controller_path);
697
698 /* Used:
699 * to get uuid to storing in bitmap metadata
700 * and 'reshape' backup-data metadata
701 * To see if a device is being re-added to an array it was part of.
702 */
703 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
704
705 /* Extract generic details from metadata. This could be details about
706 * the container, or about an individual array within the container.
707 * The determination is made either by:
708 * load_super being given a 'component' string.
709 * validate_geometry determining what to create.
710 * The info includes both array information and device information.
711 * The particular device should be:
712 * The last device added by add_to_super
713 * The device the metadata was loaded from by load_super
714 * If 'map' is present, then it is an array raid_disks long
715 * (raid_disk must already be set and correct) and it is filled
716 * with 1 for slots that are thought to be active and 0 for slots which
717 * appear to be failed/missing.
718 * *info is zeroed out before data is added.
719 */
720 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
721 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
722 /* Check if the given metadata is flagged as belonging to "this"
723 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
724 */
725 int (*match_home)(struct supertype *st, char *homehost);
726
727 /* Make one of several generic modifications to metadata
728 * prior to assembly (or other times).
729 * sparc2.2 - first bug in early 0.90 metadata
730 * super-minor - change name of 0.90 metadata
731 * summaries - 'correct' any redundant data
732 * resync - mark array as dirty to trigger a resync.
733 * uuid - set new uuid - only 0.90 or 1.x
734 * name - change the name of the array (where supported)
735 * homehost - change which host this array is tied to.
736 * devicesize - If metadata is at start of device, change recorded
737 * device size to match actual device size
738 * byteorder - swap bytes for 0.90 metadata
739 *
740 * force-one - mark that device as uptodate, not old or failed.
741 * force-array - mark array as clean if it would not otherwise
742 * assemble
743 * assemble - not sure how this is different from force-one...
744 * linear-grow-new - add a new device to a linear array, but don't
745 * change the size: so superblock still matches
746 * linear-grow-update - now change the size of the array.
747 * writemostly - set the WriteMostly1 bit in the superblock devflags
748 * readwrite - clear the WriteMostly1 bit in the superblock devflags
749 * no-bitmap - clear any record that a bitmap is present.
750 * bbl - add a bad-block-log if possible
751 * no-bbl - remove and bad-block-log is it is empty.
752 * revert-reshape - If a reshape is in progress, modify metadata so
753 * it will resume going in the opposite direction.
754 */
755 int (*update_super)(struct supertype *st, struct mdinfo *info,
756 char *update,
757 char *devname, int verbose,
758 int uuid_set, char *homehost);
759
760 /* Create new metadata for new array as described. This could
761 * be a new container, or an array in a pre-existing container.
762 * Also used to zero metadata prior to writing it to invalidate old
763 * metadata.
764 */
765 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
766 unsigned long long size, char *name,
767 char *homehost, int *uuid,
768 unsigned long long data_offset);
769
770 /* update the metadata to include new device, either at create or
771 * when hot-adding a spare.
772 */
773 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
774 int fd, char *devname,
775 unsigned long long data_offset);
776 /* update the metadata to delete a device,
777 * when hot-removing.
778 */
779 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
780
781 /* Write metadata to one device when fixing problems or adding
782 * a new device.
783 */
784 int (*store_super)(struct supertype *st, int fd);
785
786 /* Write all metadata for this array.
787 */
788 int (*write_init_super)(struct supertype *st);
789 /* Check if metadata read from one device is compatible with an array,
790 * used when assembling an array, or pseudo-assembling was with
791 * "--examine --brief"
792 * If "st" has not yet been loaded the superblock from, "tst" is
793 * moved in, otherwise the superblock in 'st' is compared with
794 * 'tst'.
795 */
796 int (*compare_super)(struct supertype *st, struct supertype *tst);
797 /* Load metadata from a single device. If 'devname' is not NULL
798 * print error messages as appropriate */
799 int (*load_super)(struct supertype *st, int fd, char *devname);
800 /* 'fd' is a 'container' md array - load array metadata from the
801 * whole container.
802 */
803 int (*load_container)(struct supertype *st, int fd, char *devname);
804 /* If 'arg' is a valid name of this metadata type, allocate and
805 * return a 'supertype' for the particular minor version */
806 struct supertype * (*match_metadata_desc)(char *arg);
807 /* If a device has the given size, and the data_offset has been
808 * requested - work out how much space is available for data.
809 * This involves adjusting for reserved space (e.g. bitmaps)
810 * and for any rounding.
811 * 'mdadm' only calls this for existing arrays where a possible
812 * spare is being added. However some super-handlers call it
813 * internally from validate_geometry when creating an array.
814 */
815 __u64 (*avail_size)(struct supertype *st, __u64 size,
816 unsigned long long data_offset);
817 /* This is similar to 'avail_size' in purpose, but is used for
818 * containers for which there is no 'component size' to compare.
819 * This reports that whole-device size which is a minimum
820 */
821 unsigned long long (*min_acceptable_spare_size)(struct supertype *st);
822 /* Find somewhere to put a bitmap - possibly auto-size it - and
823 * update the metadata to record this. The array may be newly
824 * created, in which case data_size may be updated, or it might
825 * already exist. Metadata handler can know if init_super
826 * has been called, but not write_init_super.
827 */
828 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
829 int delay, int write_behind,
830 unsigned long long size, int may_change, int major);
831 /* Seek 'fd' to start of write-intent-bitmap. Must be an
832 * md-native format bitmap
833 */
834 void (*locate_bitmap)(struct supertype *st, int fd);
835 /* if add_internal_bitmap succeeded for existing array, this
836 * writes it out.
837 */
838 int (*write_bitmap)(struct supertype *st, int fd);
839 /* Free the superblock and any other allocated data */
840 void (*free_super)(struct supertype *st);
841
842 /* validate_geometry is called with an st returned by
843 * match_metadata_desc.
844 * It should check that the geometry described is compatible with
845 * the metadata type. It will be called repeatedly as devices
846 * added to validate changing size and new devices. If there are
847 * inter-device dependencies, it should record sufficient details
848 * so these can be validated.
849 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
850 * Return value is:
851 * 1: everything is OK
852 * 0: not OK for some reason - if 'verbose', then error was reported.
853 * -1: st->sb was NULL, 'subdev' is a member of a container of this
854 * type, but array is not acceptable for some reason
855 * message was reported even if verbose is 0.
856 */
857 int (*validate_geometry)(struct supertype *st, int level, int layout,
858 int raiddisks,
859 int *chunk, unsigned long long size,
860 unsigned long long data_offset,
861 char *subdev, unsigned long long *freesize,
862 int verbose);
863
864 /* Return a linked list of 'mdinfo' structures for all arrays
865 * in the container. For non-containers, it is like
866 * getinfo_super with an allocated mdinfo.*/
867 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
868 /* query the supertype for default geometry */
869 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
870 /* Permit subarray's to be deleted from inactive containers */
871 int (*kill_subarray)(struct supertype *st); /* optional */
872 /* Permit subarray's to be modified */
873 int (*update_subarray)(struct supertype *st, char *subarray,
874 char *update, struct mddev_ident *ident); /* optional */
875 /* Check if reshape is supported for this external format.
876 * st is obtained from super_by_fd() where st->subarray[0] is
877 * initialized to indicate if reshape is being performed at the
878 * container or subarray level
879 */
880 #define APPLY_METADATA_CHANGES 1
881 #define ROLLBACK_METADATA_CHANGES 0
882
883 int (*reshape_super)(struct supertype *st,
884 unsigned long long size, int level,
885 int layout, int chunksize, int raid_disks,
886 int delta_disks, char *backup, char *dev,
887 int direction,
888 int verbose); /* optional */
889 int (*manage_reshape)( /* optional */
890 int afd, struct mdinfo *sra, struct reshape *reshape,
891 struct supertype *st, unsigned long blocks,
892 int *fds, unsigned long long *offsets,
893 int dests, int *destfd, unsigned long long *destoffsets);
894
895 /* for mdmon */
896 int (*open_new)(struct supertype *c, struct active_array *a,
897 char *inst);
898
899 /* Tell the metadata handler the current state of the array.
900 * This covers whether it is known to be consistent (no pending writes)
901 * and how far along a resync is known to have progressed
902 * (in a->resync_start).
903 * resync status is really irrelevant if the array is not consistent,
904 * but some metadata (DDF!) have a place to record the distinction.
905 * If 'consistent' is '2', then the array can mark it dirty if a
906 * resync/recovery/whatever is required, or leave it clean if not.
907 * Return value is 0 dirty (not consistent) and 1 if clean.
908 * it is only really important if consistent is passed in as '2'.
909 */
910 int (*set_array_state)(struct active_array *a, int consistent);
911
912 /* When the state of a device might have changed, we call set_disk to
913 * tell the metadata what the current state is.
914 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
915 * transitions.
916 * set_disk might be called when the state of the particular disk has
917 * not in fact changed.
918 */
919 void (*set_disk)(struct active_array *a, int n, int state);
920 void (*sync_metadata)(struct supertype *st);
921 void (*process_update)(struct supertype *st,
922 struct metadata_update *update);
923 void (*prepare_update)(struct supertype *st,
924 struct metadata_update *update);
925
926 /* activate_spare will check if the array is degraded and, if it
927 * is, try to find some spare space in the container.
928 * On success, it add appropriate updates (For process_update) to
929 * to the 'updates' list and returns a list of 'mdinfo' identifying
930 * the device, or devices as there might be multiple missing
931 * devices and multiple spares available.
932 */
933 struct mdinfo *(*activate_spare)(struct active_array *a,
934 struct metadata_update **updates);
935 /*
936 * Return statically allocated string that represents metadata specific
937 * controller domain of the disk. The domain is used in disk domain
938 * matching functions. Disks belong to the same domain if the they have
939 * the same domain from mdadm.conf and belong the same metadata domain.
940 * Returning NULL or not providing this handler means that metadata
941 * does not distinguish the differences between disks that belong to
942 * different controllers. They are in the domain specified by
943 * configuration file (mdadm.conf).
944 * In case when the metadata has the notion of domains based on disk
945 * it shall return NULL for disks that do not belong to the controller
946 * the supported domains. Such disks will form another domain and won't
947 * be mixed with supported ones.
948 */
949 const char *(*get_disk_controller_domain)(const char *path);
950
951 /* for external backup area */
952 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
953
954 int swapuuid; /* true if uuid is bigending rather than hostendian */
955 int external;
956 const char *name; /* canonical metadata name */
957 } *superlist[];
958
959 extern struct superswitch super0, super1;
960 extern struct superswitch super_imsm, super_ddf;
961 extern struct superswitch mbr, gpt;
962
963 struct metadata_update {
964 int len;
965 char *buf;
966 void *space; /* allocated space that monitor will use */
967 void **space_list; /* list of allocated spaces that monitor can
968 * use or that it returned.
969 */
970 struct metadata_update *next;
971 };
972
973 /* A supertype holds a particular collection of metadata.
974 * It identifies the metadata type by the superswitch, and the particular
975 * sub-version of that metadata type.
976 * metadata read in or created is stored in 'sb' and 'info'.
977 * There are also fields used by mdmon to track containers.
978 *
979 * A supertype may refer to:
980 * Just an array, possibly in a container
981 * A container, not identifying any particular array
982 * Info read from just one device, not yet fully describing the array/container.
983 *
984 *
985 * A supertype is created by:
986 * super_by_fd
987 * guess_super
988 * dup_super
989 */
990 struct supertype {
991 struct superswitch *ss;
992 int minor_version;
993 int max_devs;
994 char container_devnm[32]; /* devnm of container */
995 void *sb;
996 void *info;
997 void *other; /* Hack used to convert v0.90 to v1.0 */
998 unsigned long long devsize;
999 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
1000 HW/firmware related incompatability to load metadata.
1001 Used when examining metadata to display content of disk
1002 when user has no hw/firmare compatible system.
1003 */
1004 struct metadata_update *updates;
1005 struct metadata_update **update_tail;
1006
1007 /* extra stuff used by mdmon */
1008 struct active_array *arrays;
1009 int sock; /* listen to external programs */
1010 char devnm[32]; /* e.g. md0. This appears in metadata_version:
1011 * external:/md0/12
1012 */
1013 int devcnt;
1014 int retry_soon;
1015
1016 struct mdinfo *devs;
1017
1018 };
1019
1020 extern struct supertype *super_by_fd(int fd, char **subarray);
1021 enum guess_types { guess_any, guess_array, guess_partitions };
1022 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
1023 static inline struct supertype *guess_super(int fd) {
1024 return guess_super_type(fd, guess_any);
1025 }
1026 extern struct supertype *dup_super(struct supertype *st);
1027 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
1028 extern int must_be_container(int fd);
1029 extern int dev_size_from_id(dev_t id, unsigned long long *size);
1030 void wait_for(char *dev, int fd);
1031
1032 /*
1033 * Data structures for policy management.
1034 * Each device can have a policy structure that lists
1035 * various name/value pairs each possibly with a metadata associated.
1036 * The policy list is sorted by name/value/metadata
1037 */
1038 struct dev_policy {
1039 struct dev_policy *next;
1040 char *name; /* None of these strings are allocated. They are
1041 * all just references to strings which are known
1042 * to exist elsewhere.
1043 * name and metadata can be compared by address equality.
1044 */
1045 const char *metadata;
1046 const char *value;
1047 };
1048
1049 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
1050
1051 /* iterate over the sublist starting at list, having the same
1052 * 'name' as 'list', and matching the given metadata (Where
1053 * NULL matches anything
1054 */
1055 #define pol_for_each(item, list, _metadata) \
1056 for (item = list; \
1057 item && item->name == list->name; \
1058 item = item->next) \
1059 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1060 ; else
1061
1062 /*
1063 * policy records read from mdadm are largely just name-value pairs.
1064 * The names are constants, not strdupped
1065 */
1066 struct pol_rule {
1067 struct pol_rule *next;
1068 char *type; /* rule_policy or rule_part */
1069 struct rule {
1070 struct rule *next;
1071 char *name;
1072 char *value;
1073 char *dups; /* duplicates of 'value' with a partNN appended */
1074 } *rule;
1075 };
1076
1077 extern char rule_policy[], rule_part[];
1078 extern char rule_path[], rule_type[];
1079 extern char type_part[], type_disk[];
1080
1081 extern void policyline(char *line, char *type);
1082 extern void policy_add(char *type, ...);
1083 extern void policy_free(void);
1084
1085 extern struct dev_policy *path_policy(char *path, char *type);
1086 extern struct dev_policy *disk_policy(struct mdinfo *disk);
1087 extern struct dev_policy *devid_policy(int devid);
1088 extern void dev_policy_free(struct dev_policy *p);
1089
1090 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1091 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1092 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1093
1094 enum policy_action {
1095 act_default,
1096 act_include,
1097 act_re_add,
1098 act_spare, /* This only applies to bare devices */
1099 act_spare_same_slot, /* this allows non-bare devices,
1100 * but only if recent removal */
1101 act_force_spare, /* this allow non-bare devices in any case */
1102 act_err
1103 };
1104
1105 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1106 enum policy_action want);
1107 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1108 enum policy_action want);
1109
1110 struct domainlist {
1111 struct domainlist *next;
1112 const char *dom;
1113 };
1114
1115 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1116 const char *metadata);
1117 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1118 const char *metadata);
1119 extern void domainlist_add_dev(struct domainlist **dom, int devid,
1120 const char *metadata);
1121 extern void domain_free(struct domainlist *dl);
1122 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1123 const char *metadata);
1124 void domain_add(struct domainlist **domp, char *domain);
1125
1126 extern void policy_save_path(char *id_path, struct map_ent *array);
1127 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1128
1129 #if __GNUC__ < 3
1130 struct stat64;
1131 #endif
1132
1133 #define HAVE_NFTW we assume
1134 #define HAVE_FTW
1135
1136 #ifdef __UCLIBC__
1137 # include <features.h>
1138 # ifndef __UCLIBC_HAS_LFS__
1139 # define lseek64 lseek
1140 # endif
1141 # ifndef __UCLIBC_HAS_FTW__
1142 # undef HAVE_FTW
1143 # undef HAVE_NFTW
1144 # endif
1145 #endif
1146
1147 #ifdef __dietlibc__
1148 # undef HAVE_NFTW
1149 #endif
1150
1151 #if defined(__KLIBC__)
1152 # undef HAVE_NFTW
1153 # undef HAVE_FTW
1154 #endif
1155
1156 #ifndef HAVE_NFTW
1157 # define FTW_PHYS 1
1158 # ifndef HAVE_FTW
1159 struct FTW {};
1160 # endif
1161 #endif
1162
1163 #ifdef HAVE_FTW
1164 # include <ftw.h>
1165 #endif
1166
1167 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1168
1169 extern int Manage_ro(char *devname, int fd, int readonly);
1170 extern int Manage_run(char *devname, int fd, int quiet);
1171 extern int Manage_stop(char *devname, int fd, int quiet,
1172 int will_retry);
1173 extern int Manage_subdevs(char *devname, int fd,
1174 struct mddev_dev *devlist, int verbose, int test,
1175 char *update, int force);
1176 extern int autodetect(void);
1177 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1178 extern int Grow_addbitmap(char *devname, int fd,
1179 struct context *c, struct shape *s);
1180 extern int Grow_reshape(char *devname, int fd,
1181 struct mddev_dev *devlist,
1182 unsigned long long data_offset,
1183 struct context *c, struct shape *s);
1184 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1185 int *fdlist, int cnt, char *backup_file, int verbose);
1186 extern int Grow_continue(int mdfd, struct supertype *st,
1187 struct mdinfo *info, char *backup_file,
1188 int freeze_reshape);
1189
1190 extern int restore_backup(struct supertype *st,
1191 struct mdinfo *content,
1192 int working_disks,
1193 int spares,
1194 char *backup_file,
1195 int verbose);
1196 extern int Grow_continue_command(char *devname, int fd,
1197 char *backup_file, int verbose);
1198
1199 extern int Assemble(struct supertype *st, char *mddev,
1200 struct mddev_ident *ident,
1201 struct mddev_dev *devlist,
1202 struct context *c);
1203
1204 extern int Build(char *mddev, struct mddev_dev *devlist,
1205 struct shape *s, struct context *c);
1206
1207 extern int Create(struct supertype *st, char *mddev,
1208 char *name, int *uuid,
1209 int subdevs, struct mddev_dev *devlist,
1210 struct shape *s,
1211 struct context *c,
1212 unsigned long long data_offset);
1213
1214 extern int Detail(char *dev, struct context *c);
1215 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1216 extern int Query(char *dev);
1217 extern int ExamineBadblocks(char *devname, int brief, struct supertype *forcest);
1218 extern int Examine(struct mddev_dev *devlist, struct context *c,
1219 struct supertype *forcest);
1220 extern int Monitor(struct mddev_dev *devlist,
1221 char *mailaddr, char *alert_cmd,
1222 struct context *c,
1223 int daemonise, int oneshot,
1224 int dosyslog, char *pidfile, int increments,
1225 int share);
1226
1227 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1228 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1229 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1230 extern int Wait(char *dev);
1231 extern int WaitClean(char *dev, int sock, int verbose);
1232
1233 extern int Incremental(char *devname, struct context *c,
1234 struct supertype *st);
1235 extern void RebuildMap(void);
1236 extern int IncrementalScan(int verbose);
1237 extern int IncrementalRemove(char *devname, char *path, int verbose);
1238 extern int CreateBitmap(char *filename, int force, char uuid[16],
1239 unsigned long chunksize, unsigned long daemon_sleep,
1240 unsigned long write_behind,
1241 unsigned long long array_size,
1242 int major);
1243 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1244 extern int Write_rules(char *rule_name);
1245 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1246 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
1247 extern int Dump_metadata(char *dev, char *dir, struct context *c,
1248 struct supertype *st);
1249 extern int Restore_metadata(char *dev, char *dir, struct context *c,
1250 struct supertype *st, int only);
1251
1252 extern int md_get_version(int fd);
1253 extern int get_linux_version(void);
1254 extern int mdadm_version(char *version);
1255 extern unsigned long long parse_size(char *size);
1256 extern int parse_uuid(char *str, int uuid[4]);
1257 extern int parse_layout_10(char *layout);
1258 extern int parse_layout_faulty(char *layout);
1259 extern long parse_num(char *num);
1260 extern int check_ext2(int fd, char *name);
1261 extern int check_reiser(int fd, char *name);
1262 extern int check_raid(int fd, char *name);
1263 extern int check_partitions(int fd, char *dname,
1264 unsigned long long freesize,
1265 unsigned long long size);
1266
1267 extern int get_mdp_major(void);
1268 extern int dev_open(char *dev, int flags);
1269 extern int open_dev(char *devnm);
1270 extern int open_dev_flags(char *devnm, int flags);
1271 extern int open_dev_excl(char *devnm);
1272 extern int is_standard(char *dev, int *nump);
1273 extern int same_dev(char *one, char *two);
1274 extern int compare_paths (char* path1,char* path2);
1275 extern void enable_fds(int devices);
1276
1277 extern int parse_auto(char *str, char *msg, int config);
1278 extern struct mddev_ident *conf_get_ident(char *dev);
1279 extern struct mddev_dev *conf_get_devs(void);
1280 extern int conf_test_dev(char *devname);
1281 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1282 extern struct createinfo *conf_get_create_info(void);
1283 extern void set_conffile(char *file);
1284 extern char *conf_get_mailaddr(void);
1285 extern char *conf_get_mailfrom(void);
1286 extern char *conf_get_program(void);
1287 extern char *conf_get_homehost(int *require_homehostp);
1288 extern char *conf_line(FILE *file);
1289 extern char *conf_word(FILE *file, int allow_key);
1290 extern void print_quoted(char *str);
1291 extern void print_escape(char *str);
1292 extern int use_udev(void);
1293 extern unsigned long GCD(unsigned long a, unsigned long b);
1294 extern int conf_name_is_free(char *name);
1295 extern int conf_verify_devnames(struct mddev_ident *array_list);
1296 extern int devname_matches(char *name, char *match);
1297 extern struct mddev_ident *conf_match(struct supertype *st,
1298 struct mdinfo *info,
1299 char *devname,
1300 int verbose, int *rvp);
1301 extern int experimental(void);
1302
1303 extern void free_line(char *line);
1304 extern int match_oneof(char *devices, char *devname);
1305 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1306 extern const int uuid_zero[4];
1307 extern int same_uuid(int a[4], int b[4], int swapuuid);
1308 extern void copy_uuid(void *a, int b[4], int swapuuid);
1309 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1310 extern char *fname_from_uuid(struct supertype *st,
1311 struct mdinfo *info, char *buf, char sep);
1312 extern unsigned long calc_csum(void *super, int bytes);
1313 extern int enough(int level, int raid_disks, int layout, int clean,
1314 char *avail);
1315 extern int enough_fd(int fd);
1316 extern int ask(char *mesg);
1317 extern unsigned long long get_component_size(int fd);
1318 extern void remove_partitions(int fd);
1319 extern int test_partition(int fd);
1320 extern int test_partition_from_id(dev_t id);
1321 extern int get_data_disks(int level, int layout, int raid_disks);
1322 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1323 int chunksize, unsigned long long devsize);
1324 extern int flush_metadata_updates(struct supertype *st);
1325 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1326 extern int assemble_container_content(struct supertype *st, int mdfd,
1327 struct mdinfo *content,
1328 struct context *c,
1329 char *chosen_name);
1330 extern struct mdinfo *container_choose_spares(struct supertype *st,
1331 unsigned long long min_size,
1332 struct domainlist *domlist,
1333 char *spare_group,
1334 const char *metadata, int get_one);
1335 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1336 extern int add_disk(int mdfd, struct supertype *st,
1337 struct mdinfo *sra, struct mdinfo *info);
1338 extern int remove_disk(int mdfd, struct supertype *st,
1339 struct mdinfo *sra, struct mdinfo *info);
1340 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1341 unsigned long long min_recovery_start(struct mdinfo *array);
1342
1343 extern char *human_size(long long bytes);
1344 extern char *human_size_brief(long long bytes, int prefix);
1345 extern void print_r10_layout(int layout);
1346
1347 extern char *find_free_devnm(int use_partitions);
1348
1349 extern void put_md_name(char *name);
1350 extern char *devid2devnm(int devid);
1351 extern int devnm2devid(char *devnm);
1352 extern char *get_md_name(char *devnm);
1353
1354 extern char DefaultConfFile[];
1355
1356 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1357 char *chosen);
1358 /* values for 'trustworthy' */
1359 #define LOCAL 1
1360 #define LOCAL_ANY 10
1361 #define FOREIGN 2
1362 #define METADATA 3
1363 extern int open_mddev(char *dev, int report_errors);
1364 extern int open_container(int fd);
1365 extern int metadata_container_matches(char *metadata, char *devnm);
1366 extern int metadata_subdev_matches(char *metadata, char *devnm);
1367 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1368 extern int is_subarray_active(char *subarray, char *devname);
1369 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1370 extern struct superswitch *version_to_superswitch(char *vers);
1371
1372 extern int mdmon_running(char *devnm);
1373 extern int mdmon_pid(char *devnm);
1374 extern int check_env(char *name);
1375 extern __u32 random32(void);
1376 extern int start_mdmon(char *devnm);
1377
1378 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1379 struct supertype *st, unsigned long stripes,
1380 int *fds, unsigned long long *offsets,
1381 int dests, int *destfd, unsigned long long *destoffsets);
1382 void abort_reshape(struct mdinfo *sra);
1383
1384 void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0);
1385
1386 extern void fmt_devname(char *name, int num);
1387 extern char *stat2devnm(struct stat *st);
1388 extern char *fd2devnm(int fd);
1389
1390 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1391 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1392 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1393 (ROUND_UP((unsigned long)(ptr), base)))
1394
1395 static inline int is_subarray(char *vers)
1396 {
1397 /* The version string for a 'subarray' (an array in a container)
1398 * is
1399 * /containername/componentname for normal read-write arrays
1400 * -containername/componentname for arrays which mdmon must not
1401 * reconfigure. They might be read-only
1402 * or might be undergoing reshape etc.
1403 * containername is e.g. md0, md_d1
1404 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1405 */
1406 return (*vers == '/' || *vers == '-');
1407 }
1408
1409 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1410 {
1411 return &ent->metadata_version[10+strlen(container)+1];
1412 }
1413
1414 #ifdef DEBUG
1415 #define dprintf(fmt, arg...) \
1416 fprintf(stderr, fmt, ##arg)
1417 #else
1418 #define dprintf(fmt, arg...) \
1419 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1420 #endif
1421 #include <assert.h>
1422 #include <stdarg.h>
1423 static inline int xasprintf(char **strp, const char *fmt, ...) {
1424 va_list ap;
1425 int ret;
1426 va_start(ap, fmt);
1427 ret = vasprintf(strp, fmt, ap);
1428 va_end(ap);
1429 assert(ret >= 0);
1430 return ret;
1431 }
1432
1433 #define pr_err(fmt ...) fprintf(stderr, Name ": " fmt)
1434 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1435
1436 void *xmalloc(size_t len);
1437 void *xrealloc(void *ptr, size_t len);
1438 void *xcalloc(size_t num, size_t size);
1439 char *xstrdup(const char *str);
1440
1441 #define LEVEL_MULTIPATH (-4)
1442 #define LEVEL_LINEAR (-1)
1443 #define LEVEL_FAULTY (-5)
1444
1445 /* kernel module doesn't know about these */
1446 #define LEVEL_CONTAINER (-100)
1447 #define LEVEL_UNSUPPORTED (-200)
1448
1449 /* faulty stuff */
1450
1451 #define WriteTransient 0
1452 #define ReadTransient 1
1453 #define WritePersistent 2
1454 #define ReadPersistent 3
1455 #define WriteAll 4 /* doesn't go to device */
1456 #define ReadFixable 5
1457 #define Modes 6
1458
1459 #define ClearErrors 31
1460 #define ClearFaults 30
1461
1462 #define AllPersist 100 /* internal use only */
1463 #define NoPersist 101
1464
1465 #define ModeMask 0x1f
1466 #define ModeShift 5
1467
1468 #ifdef __TINYC__
1469 #undef minor
1470 #undef major
1471 #undef makedev
1472 #define minor(x) ((x)&0xff)
1473 #define major(x) (((x)>>8)&0xff)
1474 #define makedev(M,m) (((M)<<8) | (m))
1475 #endif
1476
1477 /* for raid4/5/6 */
1478 #define ALGORITHM_LEFT_ASYMMETRIC 0
1479 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1480 #define ALGORITHM_LEFT_SYMMETRIC 2
1481 #define ALGORITHM_RIGHT_SYMMETRIC 3
1482
1483 /* Define non-rotating (raid4) algorithms. These allow
1484 * conversion of raid4 to raid5.
1485 */
1486 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1487 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1488
1489 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1490 * Firstly, the exact positioning of the parity block is slightly
1491 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1492 * of DDF.
1493 * Secondly, or order of datablocks over which the Q syndrome is computed
1494 * is different.
1495 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1496 * These layouts are from the DDFv1.2 spec.
1497 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1498 * leaves RLQ=3 as 'Vendor Specific'
1499 */
1500
1501 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1502 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1503 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1504
1505 /* For every RAID5 algorithm we define a RAID6 algorithm
1506 * with exactly the same layout for data and parity, and
1507 * with the Q block always on the last device (N-1).
1508 * This allows trivial conversion from RAID5 to RAID6
1509 */
1510 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1511 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1512 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1513 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1514 #define ALGORITHM_PARITY_0_6 20
1515 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1516
1517 /* Define PATH_MAX in case we don't use glibc or standard library does
1518 * not have PATH_MAX defined. Assume max path length is 4K characters.
1519 */
1520 #ifndef PATH_MAX
1521 #define PATH_MAX 4096
1522 #endif
1523
1524 #define RESYNC_NONE -1
1525 #define RESYNC_DELAYED -2
1526 #define RESYNC_PENDING -3
1527 #define RESYNC_UNKNOWN -4
1528
1529 /* When using "GET_DISK_INFO" it isn't certain how high
1530 * we need to check. So we impose an absolute limit of
1531 * MAX_DISKS. This needs to be much more than the largest
1532 * number of devices any metadata can support. Currently
1533 * v1.x can support 1920
1534 */
1535 #define MAX_DISKS 4096
1536
1537 /* Sometimes the 'size' value passed needs to mean "Maximum".
1538 * In those cases with use MAX_SIZE
1539 */
1540 #define MAX_SIZE 1
1541
1542 /* We want to use unsigned numbers for sector counts, but need
1543 * a value for 'invalid'. Use '1'.
1544 */
1545 #define INVALID_SECTORS 1
1546 /* And another special number needed for --data_offset=variable */
1547 #define VARIABLE_OFFSET 3